A primary concern in obtaining crystalline film growth by molecular-beam epitaxy (MBE) or other vapor-phase technique is the mode of growth of the film. Three principal modes of such film growth are (1) layer-by-layer growth, referred to as a "Frank-Van der Merwe" growth mode; (2) "islanding," referred to as a "Volmer-Weber" growth mode; and (3) layer-by-layer growth to a threshold thickness, followed by islanding, referred to as a "Stranski-Krastanov" growth mode. In either growth mode which entails "islanding," the film ultimately becomes divided into domains or "islands" of crystallinity. Such a division of a film into islands of crystallinity constitutes a disruption of the long-range crystal structure desired for many applications--particularity for many applications in the field of solid-state electronics.
Surface free energies and lattice-strain energies are significant factors in determining which growth mode will be dominant when a crystallizable material is deposited on a surface of a substrate to form a film. Putting aside the matter of the lattice strain energy of the film, theoretical models of epitaxial growth suggest that the growth mode is largely determined by the surface free energy of the substrate surface (.sigma..sub.s), the surface free energy of the deposited layer (.sigma..sub.f), and an interface free energy (.sigma..sub.i). An inequality expression involving these free energies .sigma..sub.s &gt;.sigma..sub.f +.sigma..sub.i specifies a condition under which a deposited film effectively wets a substrate. When a crystallizable material deposited on a substrate wets the substrate, Frank-Van der Merwe layer-by-layer epitaxial growth may occur. If the inequality has the opposite direction, there is usually no wetting of the substrate when the crystallizable material is deposited on the substrate and Volmer-Weber immediate islanding growth tends to occur. The Stranski-Krastanov thickness-threshold islanding growth generally tends to occur when the deposited material wets the substrate, but the lattice strain energy of the resulting deposited layer is unfavorable, or when there is an added complication such as interface mixing or surface reconstruction.
It would often be desirable to fabricate a thick, structurally relaxed, substantially defect-free epitaxial layer of one element embedded in a crystalline matrix of another element. In other words, it would often be desirable to fabricate layered structures with a crystalline substrate of a first element, an embedded epitaxial layer of a second element, and a capping epitaxial layer of the first element of the substrate in which each of layers is structurally relaxed to a bulk crystalline structure and is substantially defect free.
In general, for two elements A and B, one of the elements has a lower surface free energy than the other. Consequently, if element A can be grown on element B in either a Frank-Van der Merwe layer-by-layer growth mode or a Stranski-Krastanov thickness-threshold islanding growth mode, then element B will grow on element A in a Volmer-Weber immediate islanding mode. Consequently, there is a significant barrier to the growth of an epitaxial layer of one element embedded in a crystalline matrix of the other element; that is, to the growth of layered structures of the elements in the order A/B/A or B/A/B. If the film to be embedded grows well on the substrate, then the capping layer ordinarily does not grow well on the film to be embedded. Conversely, if the capping layer were to grow well on the film to be embedded, then the film to be embedded would tend not to grow well on the substrate in the first instance.
Germanium has a lower surface free energy than that of silicon and the interface free energy .sigma..sub.i may generally be considered insignificant. Germanium growth on silicon above about 400.degree. C. follows a Stranski-Krastanov thickness-threshold islanding growth mode: at coverages below about three monolayers, layer-by-layer epitaxial growth is usually observed. At coverages corresponding to more than about three monolayers of germanium, islanding of the germanium generally occurs.
Attempts have been made to avoid such islanding by inhibiting the mobility of the germanium layer by lowering the growth temperature. However, it is generally found that films grown by low temperature processes tend to suffer from poor crystal quality and frequently from inferior electrical properties as well. Other attempts to avoid islanding of germanium on silicon have involved increasing the rate of deposition of germanium. If the germanium deposition growth rate is sufficiently high, it is possible to grow germanium layers with thicknesses which exceed the thickness normally associated with islanding. However, even using such a high-growth rate technique, it has been found that germanium films can be made only about six monolayers thick before a deterioration of the film properties is observed.
The tendency of germanium films to island after the first few monolayers can be observed experimentally using Rutherford ion backscattering analysis. In FIG. 1, 100 keV He.sup.+ Rutherford ion backscattering spectra of germanium deposited on a substrate of silicon (001) by a conventional molecular-beam-epitaxy process are plotted for the initial stages of germanium film growth at about 500.degree. C. After approximately the first three monolayers, the intensity of the leading germanium peak saturates. With further increases in coverage by germanium, the background behind the germanium peak increases, indicating that growth is restricted to islands of .gtoreq.50 .ANG. thickness. The tendency of germanium to cluster on silicon has hindered previous attempts to grow thick, substantially defect-free epitaxial films of germanium on silicon. Thick, structurally relaxed films of germanium on silicon may be grown with conventional techniques after the germanium islands have coalesced. However, such germanium films generally contain high densities of crystal defects which penetrate the entire thickness of the film.
Attempts to grow multilayer silicon-germanium structures must overcome the fundamental limitations imposed by the growth modes of the constituents. Studies of Si/Ge/Si quantum-well structures have revealed islanding of silicon capping layers as well as severe interdiffusion effects, both evidently resulting from surface energetics.